Editing Images into Reusable Layers via Compositional Edit-Prior Distillation

Wen Xue; Yong Liu; Xueheng Li; Si Wu; Yong Xu; Yipeng Sun; Yan Li; Simiu Gu

cs.CVOct 17, 2026

Editing Images into Reusable Layers via Compositional Edit-Prior Distillation

Authors: Wen Xue, Yong Liu, Xueheng Li, Si Wu, Yong Xu, Yipeng Sun, Yan Li, Simiu Gu

Organizations: South China University of Technology · Tsinghua University · University of Science and Technology of China · Peng Cheng Laboratory

Abstract

High-quality image editing requires precise modification of target regions while preserving the content of unrelated regions. Existing editing models produce flattened outputs that limit precise local manipulation and layer reuse, while decomposition methods typically optimize layer recovery separately from instruction-driven editing. To this end, we introduce LayerEditor, a unified framework that couples instruction-guided image editing with controllable generation of reusable RGBA layers. At its core, In-Context RGB-Alpha Token Modeling jointly represents the edited canvas and an ordered sequence of RGB-Alpha layers within a single token sequence, enabling semantic edits and explicit control over individual layer content. To retain the pretrained model's editing capability during adaptation to layered outputs, we propose Compositional Edit-Prior Distillation, which transfers its editing prior to both the edited canvas and the image composed from the predicted layers. To support training and evaluation, we construct LayerEditor-100K, a large-scale paired image-editing dataset with ordered RGBA layer annotations, and introduce dedicated benchmarks for layered editing. Extensive experiments demonstrate that LayerEditor maintains strong editing fidelity while generating accurate RGBA layers, supporting both flexible layer-level manipulation and multi-layer decomposition.

Figures & tables

Figure 1: LayerEditor edits images into reusable RGBA layers with user-specified layer content. The predicted layers can be independently manipulated for subsequent edits.

Figure 2: Overview of LayerEditor. LayerEditor jointly generates the edited canvas $\mathcal{D}$ and ordered RGBA stack $\mathcal{S}$ . C-EPD uses the frozen teacher to supervise both the Direct prediction and the decoded composition reconstructed from the predicted layers.

Figure 3: Overview of LayerEditor-100K. Paired editing examples and ordered RGBA annotations cover text, object, and poster-level appearance edits with $2$ – $10$ layers, providing supervision across diverse editing tasks and layer structures.

Figure 4: Qualitative comparison of layered editing. All method outputs are rendered from the predicted RGBA stacks rather than the independently generated canvases. LayerEditor follows the editing instructions while producing coherent recompositions and reusable layers.

Figure 5: Prompt-conditioned layer-content control. Both examples request six RGBA layers with specified layer content. LayerEditor performs the requested edits while assigning text rows and birds to the corresponding reusable layers.

Figure 6: Image-to-layer decomposition on Crello test set.

Appendix figures & tables6 assets

Supplementary material from the paper’s appendix.

Appendix

Parameter	Value	Parameter	Value
LoRA rank	256	Maximum pixel budget	$1024^{2}$
Flow-matching warm-up	2 epochs	Post-training budget	3 epochs
Optimizer	AdamW	Learning rate	$2\times 10^{-5}$
LR schedule	Cosine	Minimum learning rate	$5\times 10^{-6}$
LR warm-up	200 updates	Maximum C-EPD weight	0.1
Main-model batch size	64	C-EPD interval	5 updates

Appendix

Table 4: Backbone training and inference hyperparameters.

1	for $e=1,\ldots,E$ do
2	for $(I^{\mathrm{src}},u,N,Z^{*})\sim\mathcal{D}_{\mathrm{train}}$ do
3	Sample $t\sim p(t)$ and $\epsilon\sim\mathcal{N}(0,I)$ .
4	Construct $Z_{t}=(1-\sigma_{t})Z^{}+\sigma_{t}\epsilon$ and $v_{t}^{}=\epsilon-Z^{*}$ .
5	Predict $\{v_{\theta}^{X}\}_{X\in\Omega}$ from the joint noisy state and $(I^{\mathrm{src}},u,N)$ .
6	Compute the joint flow-matching loss using Equation 5 .
7	$\mathcal{L}_{\theta}\leftarrow\mathcal{L}_{\mathrm{flow}}$ . If $e>E_{\mathrm{warm}}$ , set $k\leftarrow k+1$ .

Appendix

Algorithm 1 LayerEditor training. The pretrained editing model is used as the distillation teacher.

Figure 8: LayerEditor-100K construction pipeline. Two complementary routes create editing pairs with reusable RGBA annotations from native designs and flattened editing pairs. Shared validation checks editing fidelity, layer quality, and reconstruction before export.

Figure 11: Rubric score examples. For the same editing request, progressively improved outputs receive higher scores across editing fidelity, layer separation, composition, and Alpha quality. The examples illustrate the rubric’s discriminative behavior rather than agreement with human ratings.

Figure 12: Instruction-following editing and controllable layer decomposition on Crello-Testset . Prompts specify the edit and layer count. Large output images show recompositions, and checkerboard panels show individual RGBA layers.

Figure 13: Instruction-following editing and content-controlled layer decomposition on ScaleEdit-Testset . Prompts specify the edit, layer count, and desired content separation. Large output images show recompositions, and checkerboard panels show individual RGBA layers.

Image Editing

Image editing changes selected visual content while preserving the parts of an image that should remain unchanged.

See all 302 papers

LimeCross: Context-Conditioned Layered Image Editing with Structural Consistency
May 11, 2026Ryugo Morita, Stanislav Frolov, Brian Bernhard Moser +4Text-Guided Image Editing Image Editing
RPTU Kaiserslautern-Landau & DFKI GmbH, Kaiserslautern, Germany · Faculty of Science and Engineering, Hosei University, Tokyo, Japan · EQUES, Tokyo, Japan
MRT: Masked Region Transformer for Layered Image Generation and Editing at Scale
May 26, 2026Zhicong Tang, Zhao Zhang, Jingye Chen +6Diffusion Transformer Image Generation
Canva Research
Referring Layer Decomposition
Feb 22, 2026Fangyi Chen, Yaojie Shen, Lu Xu +4Compositional Representation Learning
Intelligent Editing Team, Intelligent Creation, ByteDance

May 11, 2026J/K move · Enter open · S save

LimeCross: Context-Conditioned Layered Image Editing with Structural Consistency

Ryugo Morita, Stanislav Frolov, Brian Bernhard Moser, Ko Watanabe+3

RPTU Kaiserslautern-Landau & DFKI GmbH, Kaiserslautern, Germany · Faculty of Science and Engineering, Hosei University, Tokyo, Japan · EQUES, Tokyo, Japan

Text-Guided Image Editing Image Editing

Open paper AskPDF arXiv

Layered image assets are widely used in real-world creative workflows, enabling non-destructive iteration and flexible re-composition. Recent advances in layered image generation and decomposition synthesize or recover layered representations, yet controllable editing of layered images remains challenging. Manual editing requires careful coordination across layers to maintain consistent illumination and contact, while AI-based pipelines collapse layers into a flattened image for editing, then decompose them again, introducing background-to-foreground leakage and unstable transparency. To address these limitations, we propose LimeCross, a training-free context-conditioned layered image editing framework that edits user-selected RGBA layers according to text while keeping the remaining layers unchanged. It leverages contextual cues from other layers using a bi-stream attention mechanism to preserve cross-layer consistency, while explicitly maintaining layer integrity to prevent the contamination of edited layers. To evaluate our approach, we introduce LayerEditBench, a benchmark of 1500 layered scenes with paired source/target prompts, along with evaluation protocols that assess both edit fidelity and alpha channel stability. Extensive experiments demonstrate that LimeCross improves layer purity and composite realism over strong editing baselines, establishing context-conditioned layered editing as a principled framework for controllable generative creation.

Abstract

Image Editing

Explore similar work

LimeCross: Context-Conditioned Layered Image Editing with Structural Consistency

MRT: Masked Region Transformer for Layered Image Generation and Editing at Scale

Referring Layer Decomposition